Title:
LEARNING DEVICE, OBJECT DETECTION DEVICE, LEARNING METHOD, AND OBJECT DETECTION METHOD
Document Type and Number:
WIPO Patent Application WO/2023/017677
Kind Code:
A1
Abstract:
In the present invention, respective similarities between a correct answer region indicating a region of an object in an image and a plurality of anchor boxes set in advance in the image are acquired, and anchor boxes that exhibit not less than a specific threshold value of the similarities are selected for the correct answer region, from among the plurality of anchor boxes. Training to obtain a neural network model for detecting an object is performed on the basis of the correct answer region and the selected anchor boxes. In a case where the upper limit number of anchor boxes have been selected for the correct answer region, the upper limit number of anchor boxes corresponding to the correct answer region is changed on the basis of the similarities acquired for the anchor boxes selected for the correct answer region.
Inventors:
IZAWA HIDETOSHI (JP)
SHIMODAIRA HAYATO (JP)
SHIMODAIRA HAYATO (JP)
Application Number:
PCT/JP2022/024325
Publication Date:
February 16, 2023
Filing Date:
June 17, 2022
Export Citation:
Assignee:
CANON KK (JP)
International Classes:
G06T7/00; G06V10/82
Domestic Patent References:
| WO2020214580A1 | 2020-10-22 |
Foreign References:
| CN110991468A | 2020-04-10 |
Other References:
REDMON JOSEPH; FARHADI ALI: "YOLO9000: Better, Faster, Stronger", 2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), IEEE COMPUTER SOCIETY, US, 21 July 2017 (2017-07-21), US , pages 6517 - 6525, XP033250016, ISSN: 1063-6919, DOI: 10.1109/CVPR.2017.690
YANG XUE; YANG JIRUI; YAN JUNCHI; ZHANG YUE; ZHANG TENGFEI; GUO ZHI; SUN XIAN; FU KUN: "SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects", 2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), IEEE, 27 October 2019 (2019-10-27), pages 8231 - 8240, XP033723181, DOI: 10.1109/ICCV.2019.00832
YANG XUE; YANG JIRUI; YAN JUNCHI; ZHANG YUE; ZHANG TENGFEI; GUO ZHI; SUN XIAN; FU KUN: "SCRDet: Towards More Robust Detection for Small, Cluttered and Rotated Objects", 2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), IEEE, 27 October 2019 (2019-10-27), pages 8231 - 8240, XP033723181, DOI: 10.1109/ICCV.2019.00832
Attorney, Agent or Firm:
OHTSUKA PATENT OFFICE, P.C. (JP)
Download PDF:
Previous Patent: FILTER DEVICE, AND HIGH-FREQUENCY FRONT END CIRCUIT PROVIDED WITH SAME
Next Patent: ELECTRONIC DEVICE AND METHOD FOR PRODUCING SAME
Next Patent: ELECTRONIC DEVICE AND METHOD FOR PRODUCING SAME